Termination of Estimation


  1. After I run a maximum likelihood estimation using fmincon, I got the following message:

Optimization terminated: magnitude of search direction less than 2*options.TolX
and maximum constraint violation is less than options.TolCon.

In the meanwhile, the first-order optimality is still a large number.
Is this a good termination? Shoudn’t the first-order optimality be close to zero?

  1. If the t value for some estimated parameters are very small, does that mean these parameters cannot be accurately estimated with data? The likelihood is very flat over these parameters?

I also noticed that the estimated value will not deviate a lot from the initial values for these parameters. Does that mean data do not have enough information to estimate these parameters?



  1. This message seems to indicate that the value of the parameters aren’t moving enough anymore but that an optimum hasn’t been reached.

  2. Remember that the t-stat is computed for the null hypothesis that a given parameter is equal to zero. The t-stat can only be used to decide that test.